Search Results

Visage Breast Density is a software application intended for use with compatible full field digital mammography and digital breast tomosynthesis systems. Visage Breast Density assesses breast density from a mammography study and provides an ACR BI-RADS Atlas 5th Edition breast density category to aid radiologists in the assessment of breast tissue composition. Visage Breast Density produces adjunctive information. It is not a diagnostic aid.

Visage Breast Density is a software application that assesses breast density from a mammography study and provides a density category A, B, C, or D according to the ACR BI-RADS Atlas 5th Edition to aid radiologists in the assessment of breast tissue composition.

Visage Breast Density employs a convolutional network (CNN) for the automatic classification of breast density. The CNN has been trained on a large database of mammography exams. When applied to a mammography image, the CNN computes four likelihoods corresponding to the four breast density categories. The classifications of the individual images are merged into a general classification of the mammography study.

Visage Breast Density is designed as an add-on module to the Visage 7 product for distributing, viewing, processing, and archiving medical images. The assessment of breast density is performed from mammography studies stored on the Visage 7 server. The resulting breast density classification is displayed by the Visage 7 client on a computer monitor and stored in the database on the Visage 7 server.

Here's a breakdown of the acceptance criteria and study proving device efficacy for Visage Breast Density, based on the provided document:

1. Table of Acceptance Criteria and Reported Device Performance

The document states that the acceptance criteria were defined by comparing the performance of Visage Breast Density to that of the predicate device, PowerLook Density Assessment. Specifically, the acceptance criteria are implicit in the statement: "Visage Breast Density achieved similar accuracies per category and similar total accuracies compared to the predicate device."

While explicit numerical acceptance criteria (e.g., "accuracy must be >= X%") are not provided, the "reported device performance" is the claim of "similar accuracies per category and similar total accuracies compared to the predicate device."

Therefore, the table would look like this:

2. Sample Size Used for the Test Set and Data Provenance

• Sample Size:
  • Test Set 1: 500 studies
  • Test Set 2: 700 studies
  • Total Test Set Size: 1200 studies
• Data Provenance: "two different sites." The country of origin is not explicitly stated, but given the company's German location (Visage Imaging GmbH, Berlin, Germany), it's plausible the data is from Europe, potentially Germany. The document does not specify if the data was retrospective or prospective, but it's common for such studies to use retrospective data.

3. Number of Experts Used to Establish the Ground Truth for the Test Set and Qualifications of Those Experts

• Number of Experts: Three board-certified radiologists per site were used to establish the consensus ground truth. Since there were two sites, it implies a total of 6 unique radiologists (3 per site).
• Qualifications: "Three board certified radiologists with MQSA qualification per site." MQSA (Mammography Quality Standards Act) qualification is a US standard, which might suggest US sites or radiologists with equivalent qualifications.

4. Adjudication Method for the Test Set

The adjudication method was consensus. "the consensus of the three reviewers was determined for each study." This implies that the three radiologists reviewed each case, and their agreement (or a process to resolve disagreement) led to the final ground truth label. A common consensus method is a majority vote (e.g., 2 out of 3 agree).

5. If a Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study was done, If so, what was the effect size of how much human readers improve with AI vs without AI assistance

No, a Multi-Reader Multi-Case (MRMC) comparative effectiveness study was not conducted to assess how human readers improve with AI assistance. The study focuses on the standalone performance of the AI by comparing its output to a human-established ground truth. The device is described as providing "adjunctive information," not as a diagnostic aid that would directly assist human reader performance.

6. If a Standalone (i.e., algorithm only without human-in-the-loop performance) was done

Yes, a standalone performance evaluation of the algorithm was done. The study assessed "The predicted breast density category of Visage Breast Density... related to the ground truth from the clinical reports and the consensus of the three reviewers." This directly measures the algorithm's performance independent of human input during the assessment process itself.

7. The Type of Ground Truth Used

The ground truth used was a combination of:

• Expert Consensus: "consensus of the three reviewers" (radiologists).
• Clinical Reports: "ground truth from the clinical reports."

This suggests the radiologists reviewed the clinical reports and then formed a consensus, or perhaps the clinical reports served as an initial "gold standard" which was then validated/adjudicated by the expert radiologists.

8. The Sample Size for the Training Set

The sample size for the training set is not explicitly stated. The document only mentions: "The CNN has been trained on a large database of mammography exams."

9. How the Ground Truth for the Training Set was Established

The document does not explicitly describe how the ground truth for the training set was established. It only states that the CNN was "trained on a large database of mammography exams." Typically, for such training, the ground truth would also be established by expert radiologists, likely with similar methods as the test set (consensus or single expert review).

Ask a specific question about this device